The integrin alphav beta3 increases cellular stiffness and cytoskeletal remodeling dynamics to facilitate cancer cell invasion

The process of cancer cell invasion through the extracellular matrix (ECM) of connective tissue plays a prominent role in tumor progression and is based fundamentally on biomechanics. Cancer cell invasion usually requires cell adhesion to the ECM through the cell-matrix adhesion receptors integrins. The expression of the αvβ3 integrin is increased in several tumor types and is consistently associated with increased metastasis formation in patients. The hypothesis was that the αvβ3 integrin expression increases the invasiveness of cancer cells through increased cellular stiffness, and increased cytoskeletal remodeling dynamics. Here, the invasion of cancer cells with different αvβ3 integrin expression levels into dense three-dimensional (3D) ECMs has been studied. Using a cell sorter, two subcell lines expressing either high or low amounts of αvβ3 integrins (αvβ3high or αvβ3low cells, respectively) have been isolated from parental MDA-MB-231 breast cancer cells. αvβ3high cells showed a threefold increased cell invasion compared to αvβ3low cells. Similar results were obtained for A375 melanoma, 786-O kidney and T24 bladder carcinoma cells, and cells in which the β3 integrin subunit was knocked down using specific siRNA. To investigate whether contractile forces are essential for αvβ3 integrin-mediated increased cellular stiffness and subsequently enhanced cancer cell invasion, invasion assays were performed in the presence of myosin light chain kinase inhibitor ML-7 and Rho kinase inhibitor Y27632. Indeed, cancer cell invasiveness was reduced after addition of ML-7 and Y27632 in αvβ3high cells but not in αvβ3low cells. Moreover, after addition of the contractility enhancer calyculin A, an increase in pre-stress in αvβ3low cells was observed, which enhanced cellular invasiveness. In addition, inhibition of the Src kinase, STAT3 or Rac1 strongly reduced the invasiveness of αvβ3high cells, whereas the invasiveness of β3 specific knock-down cells and αvβ3low cells was not altered. In summary, these results suggest that the αvβ3 integrin enhances cancer cell invasion through increased cellular stiffness and enhanced cytoskeletal remodeling dynamics, which enables the cells to generate and transmit contractile forces to overcome the steric hindrance of 3D ECMs.

cancer cell invasion, invasion assays were performed in the presence of myosin light chain kinase inhibitor ML-7 and Rho kinase inhibitor Y27632. Indeed, cancer cell invasiveness was reduced after addition of ML-7 and Y27632 in αvβ3 high cells but not in αvβ3 low cells. Moreover, after addition of the contractility enhancer calyculin A, an increase in pre-stress in αvβ3 low cells was observed, which enhanced cellular invasiveness. In addition, inhibition of the Src kinase, STAT3 or Rac1 strongly reduced the invasiveness of αvβ3 high cells, whereas the invasiveness of β3 specific knock-down cells and αvβ3 low cells was not altered. In summary, these results suggest that the αvβ3 integrin enhances cancer cell invasion through increased cellular stiffness and enhanced cytoskeletal remodeling dynamics, which enables the cells to generate and transmit contractile forces to overcome the steric hindrance of 3D ECMs.

Introduction
In malignant tumor progression, the invasion of cancer cells through the extracellular matrix (ECM) of connective tissue plays a prominent role for metastasis formation [1][2][3]. This process of metastasis formation consists of multiple steps involving the detachment of invasive cells from the primary tumor mass, cancer cell migration into the surrounding tissue and transmigration (intravasation) through blood or lymph vessels possibly with the help of macrophages and their extravasation into the target organs [4][5][6][7].
The metastatic process requires dynamic biomechanical alterations of the homotypic intercellular interactions of cancer cells and heterotypic intercellular interactions of cancer cells and other cell types. Additionally, these interactions of cancer cells are modulated by components of the ECM [8]. Indeed, several matrix-cell surface adhesion receptors such as cadherins and integrins have been identified to act as either as negative or positive regulators of cancer cell invasion and metastasis [9][10][11][12].
Cell-surface expressed integrins couple the ECM to cytoskeletal actin microfilaments. This coupling is an initiation signal for focal adhesion proteins to cluster underneath the cell membrane to focal contacts and focal adhesions [13]. Thus, focal adhesion proteins such as vinculin and focal adhesion kinase are critical for the process of cellular motility into ECM [14][15][16][17]. In focal adhesions, cell surface transmembrane receptors of the integrin family promote the extracellular interaction with ECM ligands and couple the microenvironment with the cytoskeletal actin-microfilament system [18][19]. Indeed, focal adhesion sites, or rather their components, not only function to facilitate cell-matrix adhesion or cell motility, but also serve as a membrane anchor for the cytoskeleton and transduction of biochemical and biomechanical signals from the ECM to the cell and vice versa [18][19][20][21][22].
Moreover, the β3 integrin subunit is a component of the αvβ3 integrin heterodimer, which plays a prominent role in blood vessel formation/angiogenesis and hence promotes tumor growth [23][24][25][26].
As most of the previous studies on cancer cell migration examined only individual cell lines from specific tumor types, the results are still difficult to generalize. In particular, it is unknown whether molecular determinants of cancer cell migration exist which are common to different tumor cell types. In order to approach this problem the study analyzes cancer cell lines from breast, skin, kidney and bladder for their ability to invade into dense three-dimensional extracellular matrices (3D ECMs) in the absence and presence of high and low αvβ3 integrin expression in vitro.
The aim of this study was to analyze the role of the αvβ3 integrins for cancer cell invasion under controlled in vitro conditions, and to characterize the biomechanical invasion strategy that is activated by αvβ3 integrins. The 2.4 mg ml −1 synthetic 3D ECMs with subcellularsized pores were used for the invasion assays [37][38][39]. The invasiveness of cancer cells in such a system depends basically on a balanced regulation of biomechanical processes including cell adhesion and de-adhesion [40], cytoskeletal remodeling [41], protrusive force generation [40,42] and matrix properties such as stiffness, pore size, ECM protein composition and enzymatic degradation [43].
In particular, it was investigated whether the expression of αvβ3 integrins facilitates 3D ECM invasion through enhanced cellular stiffness and increased cytoskeletal remodeling dynamics, as needed to overcome the steric hindrance of dense 3D ECMs. To address this issue, cancer cells with high and low endogenous αvβ3 integrin expression were isolated using flow cytometry and used for invasion and biomechanical analysis. αvβ3 high cancer cells displayed increased invasiveness into 3D ECMs compared to αvβ3 low cells. Consistently, knock-down of the β3 integrin subunit in αvβ3 high cells decreased cancer cell invasion into 3D ECMs. The αvβ3 integrin specificity of the invasion-enhancing effect was analyzed systematically by measuring cell cytoskeletal remodeling and cellular stiffness. Moreover, it was explored whether αvβ3 integrin mediated invasiveness in 3D ECMs is ROCK, MLCK, Src, STAT3 and Rac1 signaling dependent. Taken together, this study reveals that αvβ3 integrins contribute substantially to the invasiveness of cancer cells by promoting the signaling for the transmission and generation of contractile forces and hence, increasing cellular stiffness.

High αvβ3 integrin expression leads to enhanced cell invasion
To investigate the effect of the αvβ3 integrin expression on invasion, two subcell lines were isolated from the parental breast cancer line MDA-MB-231 that expressed either high or low amounts of αvβ3 integrins on their cell surface ( figure 1(A)). In the following, these subcell lines are referred to as αvβ3 low and αvβ3 high cells. Between these subcell lines, the difference in the expression of αvβ3 integrin was 46-fold, which is shown as mean fluorescence intensities (MFIs) (figure 1(A)). Using cytofluorometry, the αvβ3 integrin expression levels were confirmed to be stable during culture for more than 50 passages (data not shown). All experiments in this study were performed on cells derived from a single isolation that had been obtained by cell sorting with respect to αvβ3 integrin subunit expression. In independent experiments, the sorting of parental cells in total three times over the course of 2 years was repeated, and each time stable αvβ3 low and αvβ3 high phenotypes with similarly high differences in the invasion behavior were established (data not shown). This finding confirms that the αvβ3 low and αvβ3 high phenotypes can be obtained reproducibly. In addition, the expression profiles of other relevant integrins such as α1, α2, α5 and β1 integrin subunits were analyzed on the cell surface of αvβ3 low and αvβ3 high cells as well as parental MDA-MB-231 cells (figure S1, see the supplementary data, available from stacks.iop.org/NJP/15/015003/mmedia). Between αvβ3 low and αvβ3 high cells there was no difference in integrin cell surface expression observed in all other integrins tested, whereas the parental MDA-MB-231 showed significantly increased levels of α5 integrin subunit expression compared to both αvβ3 low and αvβ3 high cells (figure S1). The expression of αvβ3 integrin on the cell surface of parental MDA-MB-231 is between the two low and high αvβ3 integrin subtypes (figure S2).
The percentage of cells that were able to invade into a 3D ECM significantly was higher for αvβ3 high than for αvβ3 low cells ( figure 1(B)). In addition, the invasion profile (cumulative probability) of the invasive cells showed that αvβ3 high cells invaded deeper into the 3D ECM (figure 1(C)). To investigate whether the effect of the αvβ3 expression on invasiveness is cancer cell-type specific, αvβ3 high and αvβ3 low cells were isolated from A375 melanoma cells (17-fold  difference between αvβ3 high and αvβ3 low , figure 1(D)), 786-O human kidney carcinoma cells (8-fold difference between αvβ3 high and αvβ3 low , figure 1(G)) as well as αvβ3 high and αvβ3 low cells from T24 bladder carcinoma cells (6-fold difference between αvβ3 high and αvβ3 low , figure 1(J)). The cell invasiveness of αvβ3 high cells derived from A375, 786-O and T24 cells was higher than that of αvβ3 low cells indicated by increased numbers of invasive cells (figures 1(E), (H) and (K)) and the invasion profiles (figures 1(F), (J) and (L)) showing that the αvβ3 high cells invaded deeper into 3D ECMs. These results confirm that the cell invasiveness increases with integrin αvβ3 expression levels in several cancer cell-types.

High αvβ3 integrin expression leads to reduced cell motility in two dimensions (2D)
To investigate the motility of the αvβ3 integrin expression on cell motility in a second alternative approach to cross-validate the invasion results in 3D and to compare the results with wellknown two-dimensional (2D) migration assays, the movement of the subcell lines αvβ3 low and αvβ3 high cancer cells were analyzed after 8 h of cell adhesion on 50 µg ml −1 fibronectin-coated glass slides for 2-6 h in a 2D microenvironment. The αvβ3 low cells with reduced αvβ3 integrin expression showed significantly faster migration speeds (figure 2(A)) and reduced migration persistence of movement (figure 2(B)) compared to the αvβ3 high cells with high αvβ3 integrin expression. As the results of 2D migration assays in this study and in other studies [44] have been reported to be contrary to the results obtained using a 3D microenvironment, the invasiveness of the subcell lines was analyzed in the following using a 3D ECM migration assay. The latter method of investigating cell motility in a tissue microenvironment seems to more reliable and mimics the tumor microenvironment better than a planar substrate, on which the surfaces of cells do not have contact with the substrate/microenvironment in all directions.

β3 integrin subunit knock-down decreases invasiveness into three-dimensional extracellular matrices (3D-ECMs)
To investigate the effect of β3 integrin subunit on cell invasion, the β3 integrin subunit was knocked-down in αvβ3 high cells derived from parental MDA-MB-231 breast cancer cells by using specific β3 siRNA. Typically representative invasive cells of αvβ3 high cells treated with control siRNA (left) or specific β3 siRNA (siβ3-1 (middle) and siβ3-2 (right)) are shown in figure 3(A). Specific knock-down of β3 in αvβ3 high cells was over 98% with residual of 1.6% ± 0.29(n = 3) expression of the αvβ3 integrin for siβ3-1 and was over 95% with residual expression of 4.4% ± 0.33(n = 3) for siβ3-2 after 2 days (figure 3(B)). The percentage of cells that were able to invade into a 3D collagen matrix was higher for β3 expressing cells (control siRNA treated cells) compared to β3 knock-down cells (β3 siRNA treated cells; two different specific β3 siRNAs: siβ3-1 and siβ3-2) (figure 3(C)). In addition, the invasion profile (cumulative probability) of the invasive cells showed that control siRNA treated cells invaded deeper into the ECM compared to β3 knock-down cells (figure 3(D)). The invasion profiles of β3 knock-down cells reveal that these cells only invade less than 50 µm (figure 3(D)) and their average invasion depth was reduced from 173.56 ± 5.6 (n = 356) to 15.4 ± 1.4(n = 111) (figure 3(E)). These results indicate that β3 integrin subunit expression leads to enhanced cancer cell invasion in 3D collagen matrices. In addition, the parental MDA-MB-231 cells have also been treated with control siRNA and specific β3 integrin (siβ3-1) siRNA and analyzed for their invasiveness into 3D ECMs (figure S3(A), see the supplementary data, available from stacks.iop.org/NJP/15/015003/mmedia). The knock-down efficiency was at least over 90%(n = 4). Indeed, the percentage of invasive cells (figure S3(B)) and their invasion depths as shown in the invasion profiles is significantly reduced in β3 integrin knock-down cells compared to control siRNA treated wildtype cells (figure S3(C)). These results indicate that also parental MDA-MB-231 cells can migrate via an αvβ3 integrin facilitated pathway into dense 3D ECMs.

Effect of the αvβ3 expression on the stiffness of cancer cells
Several mechanical properties of cancer cells can determine the efficiency of cancer cell invasion into dense 3D ECMs and hence support metastasis formation. Among them are cellular stiffness, which may determine contractile force transmission and generation, as well as cytoskeletal remodeling dynamics, which reorganizes the cell's acto-myosin cytoskeleton  (middle) and siβ3-2 (right) for 2 days. In each histogram, left curves are isotype controls and filled gray curves show integrin expression. One representative experiment out of at least three is shown. The bar graphs contain MFI (mean ± SD) values (n = 3). * * p < 0.01. (C) A higher percentage (mean values ± SE) of control siRNA treated αvβ3 high cells (black) invaded into 3D ECMs compared to siβ3-1 (light gray) and siβ3-2 (dark gray) treated αvβ3 high cells after 3 days. * * * p < 0.001. (D) Invasion profiles showed that control siRNA treated αvβ3 high cells migrated deeper into 3D collagen matrices compared to siβ3-1 (light gray) and siβ3-2 (dark gray) treated αvβ3 high 24αvβ3 low cells (dark gray) after 3 days.
(E) Invasion depth (mean values ± SE) is increased in control siRNA treated αvβ3 high cells compared to specific b3 integrin siRNA siβ3-1 (light gray) and siβ3-2 (dark gray) treated αvβ3 high cells.
including stress fibers and the turnover of focal adhesions connecting the external ECM to the cell's cytoskeletal scaffold. Both mechanical properties may support αvβ3 integrin facilitated cancer cell invasiveness into 3D ECMs. In order to analyze whether the expression of the αvβ3 integrin affects cellular adhesion strength or mechanical stiffness, αvβ3 high cells and αvβ3 low cells were measured using magnetic tweezer microrheology (figure 4). External forces of up to 10 nN were applied to super-paramagnetic beads coated with fibronectin (also a component of the 3D ECMs, figure S4-available from stacks.iop.org/NJP/15/015003/mmedia). Cancer cells secrete fibronectin into the medium and it is also present in fetal calf serum (FCS; one batch used for all experiments; figure S4). Thus, the fibronectin beads were bound to αvβ3 high cells (figure 4(A)) and αvβ3 low cells ( figure 4(B)). The displacement of the bound beads during a step-wise increased application of force (creep measurement) followed a power law [39].
The stiffness measurements (averaged over all forces from 0.5 to 10 nN) of the αvβ3 high cells and αvβ3 low cells showed that αvβ3 high cells have higher stiffness values expressed as mean values (figure 4(C)). These results indicate that the αvβ3 integrin alters cell stiffness and stressstiffening of cancer cells. To analyze whether there are differences independent of alterations between the two cell lines, we used a β3-specific siRNA approach. The β3-specific siRNA reduced the MFI on αvβ3 high cells after 2 days (figure 4(E)). These stiffness results showed a pronounced reduction in αvβ3 high cells that had been treated with the β3-specific siRNA siβ3-1 for 2 days as well as siβ3-2 compared to control siRNA treated cells (figure 4(F)). The stiffness results for the αvβ3 high cells and αvβ3 low cells and knock-down experiments suggest that contractile forces may play a role in αvβ3 facilitated cell invasion. To investigate whether the difference in stiffness is a result of fibronectin engagement with the αvβ3 integrin or also a result of collagen engagement with the αvβ3 integrin, collagen type I coated beads were used to measure cancer cell stiffness (figures 4(H) and (I)). The difference in stiffness (figure 4(H)) was also significantly present when using collagen coated beads instead of fibronectin-coated beads (figure 4(C)). This indicates that the difference in stiffness between αvβ3 high and αvβ3 low cells is a result of fibronectin/collagen engagement with the αvβ3 integrin. These findings may indicate that the αvβ3 integrin facilitated invasiveness is stiffness dependent.

Effect of the αvβ3 expression on cytoskeletal remodeling dynamics
Cancer cell invasion and metastasis involves dynamic filamentous actin cytoskeletal remodeling and assembly/disassembly of focal adhesion sites. In more detail, cytoskeletal remodeling means activation of ERK1/2, Src and focal adhesion kinase signaling pathways. These alterations include formation and dissolution of stress fibers, dynamic actin accumulation at the cellular periphery and formation of lamellipodia and filopodia. Many biomechanical parameters such as adhesion/de-adhesion, contractile forces, cellular stiffness, cytoskeletal remodeling dynamics (cellular fluidity) and matrix degradation through secreted enzymes determine the migration speed and the invasiveness of cancer cells. Indeed, cellular stiffness and cytoskeletal remodeling dynamics are related, but it is still unclear whether they are positively or negatively correlated. During cell invasion into a dense 3D ECM with pore sizes smaller than the cell's diameter, the invasive cell might change its shape and restructure its cytoskeleton to move forward in these dense 3D ECMs. The dynamics of cytoskeletal remodeling processes can also be measured with magnetic tweezer microrheology. The power-law exponent b characterizes the visco-elastic response of cancer cells and assumes typical values between 0 for elastic solid materials and 1 for viscous fluid materials. The b-values were significantly increased in αvβ3 high  The stiffness (mean values ± SE) of αvβ3 high cells and αvβ3 low cells as well as (F) of αvβ3 high cells transfected with control siRNA (control) and two β3 integrin subunit specific (siβ3-1 and siβ3-2) siRNAs was measured after force application to fibronectin-coated beads using magnetic tweezers. (E) Flow cytometric analysis (MFI as mean ± SD, n = 3) of αvβ3 expression on the cell surface of αvβ3 high cells transfected with control siRNA (right histogram) or two β3 integrin subunit specific (siβ3-1 and siβ3-2; left and middle histogram, respectively) siRNAs was measured, respectively. (D) Creep exponent b (cell fluidity and cytoskeletal remodeling dynamics; mean values ± SE) of αvβ3 high cells and αvβ3 low cells as well as (G) αvβ3 high cells transfected with control siRNA (control) and two β3 integrin subunit specific (siβ3-1 and siβ3-2) siRNAs was also determined after force application to fibronectin-coated beads using magnetic tweezers. The values are expressed as mean ± SE. 87-105 cells were measured for each condition. (H) The stiffness (mean values ± SE) of αvβ3 high cells and αvβ3 low cells was measured after force application to collagen type I (collagen)-coated beads using magnetic tweezers. (I) Creep exponent b (mean values ± SE) of αvβ3 high cells and αvβ3 low was also determined after force application to collagen-coated beads using magnetic tweezers. * * * p < 0.001. cells at all external forces applied to fibronectin-coated bound beads compared to αvβ3 low cells, indicating that these cells were more fluid-like and that the cytoskeletal remodeling dynamics was increased in αvβ3 high cells ( figure 4(D)). These results were confirmed by αvβ3 high cells that had been treated with two β3-specific siRNAs (siβ3-1 and siβ3-2) for 2 days, as these cells displayed significantly increased cellular fluidity and hence, increased cytoskeletal remodeling dynamics compared to control siRNA treated cells ( figure 4(G)). To analyze whether the collagen engagement of αvβ3 integrins lead to similar results of cellular fluidity and cytoskeletal remodeling dynamics, the b-value measurements were performed with collagen-coated beads. The b-values were significantly increased in αvβ3 high cells at all external forces applied to collagen-coated bound beads compared to αvβ3 low cells, indicating that these cells were more fluid-like (figure 4(I)) and that the cytoskeletal remodeling dynamics was increased in αvβ3 high cells. Taken together, the fibronectin/collagen engagement of the αvβ3 integrin is critical for determining cytoskeletal remodeling dynamics (cellular fluidity). These findings suggest that the αvβ3 integrin facilitated invasiveness possibly depends on the cytoskeletal remodeling dynamics.

The αvβ3 facilitated cell invasion depends on the transmission and generation of contractile forces
To investigate whether contractile forces are essential for αvβ3 integrin-mediated increased cellular stiffness and subsequently enhanced cancer cell invasion, invasion assays were performed in the presence of myosin light chain kinase inhibitor ML-7 and Rho kinase inhibitor Y27632. Indeed, cancer cell invasiveness was reduced after addition of ML-7 and Y27632 in αvβ3 high cells, but not in αvβ3 low cells (figures 5(A)-(D)). In particular the number of invasive cells of αvβ3 high cells was significantly reduced as well as their invasion depths after addition of ML-7 and Y27632 (figures 5(A) and (C)). Moreover, after addition of the contractility inducer calyculin A (Cal A), an increase in pre-stress in αvβ3 low cells was observed, which enhanced cellular invasiveness (figures 5(B) and (D)), whereas no significant change occurred in αvβ3 high cells regarding numbers of invasive cells or invasion depths (figures 5(A) and (C)). In addition, the effect of the contractile force inhibitors and the inducer of contractile forces on cellular stiffness were analyzed by adding these drugs to αvβ3 high and αvβ3 low cells using magnetic tweezer method. Indeed, ML-7 and Y27632, which both reduced contractile forces in αvβ3 high cells, also decreased the cellular stiffness, whereas the inducer of contractile forces Cal A had no further effect on the cellular stiffness of αvβ3 high cells ( figure 5(E)). In contrast, both inhibitors, ML-7 and Y27632, could not further reduce the stiffness of αvβ3 low cells, whereas the inducer of contractile forces Cal A was able to increase cellular stiffness of αvβ3 low cells ( figure 5(F)). These results suggest that the cellular stiffness regulates the transmission or generation of contractile forces, which are needed to overcome the hindrances of dense 3D ECMs. These findings may indicate that the αvβ3 integrin facilitated invasiveness depends on cellular stiffness which subsequently enables the cells to transmit and generate contractile forces.

The αvβ3 facilitated cell invasion is inhibited by the Src, STAT3 and RAC1 inhibitors
The addition of a Src kinase inhibitor (Src inh) to αvβ3 high cells (figure 6(A)) and αvβ3 high cells treated with control siRNA (figure 6(F)) reduced significantly the percentage of invasive cells into 3D ECMs, whereas it has no effect on the invasiveness of αvβ3 low cells (figure 6(C)) Fluores. intensity   ECMs, whereas it has no effect on the invasiveness of αvβ3 low cells (figure 6(C)) or αvβ3 high cells treated with specific β3 siRNA (siβ3-1) (figure 6(H)). Expression of the αvβ3 integrin on αvβ3 high cells treated with control siRNA or β3 integrin specific siRNA (figure 6(E)). Additionally, the invasion profiles show that the invasion depth is also reduced after addition of the Src, STAT3 and Rac1 inhibitors in αvβ3 high cells ( figure 6(B)) and αvβ3 high cells treated with control siRNA (figure 6(G)), but not in αvβ3 low cells (figure 6(D)) or αvβ3 high cells treated with specific β3 siRNA (figure 6(I)). Taken together, inhibition of the src kinase, STAT3 or Rac1 strongly reduced the invasiveness of αvβ3 high cells (figures 6(A) and (B)), whereas the invasiveness of β3 integrin subunit knock-down cells (figures 6(H) and (I)) and αvβ3 low cells were not altered (figures 6(C) and (D)). These results suggest that the increased invasiveness may be facilitated by a Src kinase, STAT3 and Rac1 pathway. In addition, the effect of these inhibitors on cellular stiffness was analyzed in αvβ3 high and αvβ3 low cells using magnetic tweezers. The Src, STAT3 and Rac1 inhibitors reduced the cellular stiffness in αvβ3 high cells (figure 6(J)), but had no effect on the stiffness of αvβ3 low cells (figure 6(K)). These results demonstrate that the αvβ3 integrin-facilitated invasiveness of cancer cells depends on the cellular stiffness.

Discussion
Previously, the integrin αvβ3 has been reported to be involved in the malignant progression of neoplasms involving tumor growth and metastasis formation [29][30][31][32][33][34]. Consistent with these studies we showed that high αvβ3 integrin expression increased MDA-MB-231 breast carcinoma cell invasiveness into 3D ECMs. In contrast, using 2D migration assays it has been reported that the αvβ3 integrin reduces the motility on ECM-protein ligand coated substrates [44]. Indeed, we confirmed that the motility of αvβ3 high and αvβ3 low cells was altered in terms of migration speed or persistence of migration. In particular, the αvβ3 low cells had a significantly higher migration speed, but migrated less persistently compared to αvβ3 high cells. These results demonstrate that the αvβ3 integrin decreased the invasiveness on 2D planar substrates. Hence, these findings indicate that the effect of the αvβ3 integrin on cancer cell motility depends highly on the invasion microenvironment, in particular on 2D or 3D microenvironment. In particular, the dimensionality of the migration assay performed clearly affects the cellular motility and one should consider before a migration assay is used how this may influence the migrations results and mimic the natural cellular microenvironment.
Additionally, we showed that high endogenous αvβ3 integrin expressing A375 (melanoma), T24 (bladder) and 786-O kidney cancer cell subclones increased the invasiveness into 3D ECMs compared to low endogenous αvβ3 integrin expressing subclones. These results suggest that the αvβ3 integrin facilitated invasiveness is not restricted to one cancer cell type, as it is transferable to other cancer types. The invasion of cancer cells into their microenvironment such as connective tissue is a multistep event and depends on mechanical and biochemical properties of the microenvironment including ECM proteins [37,[43][44][45] and embedded cells [46] as well as cancer cells [47]. Here, we demonstrate that the high expression of the αvβ3 integrin enhances invasiveness of cancer cells into 3D ECMs through increased cellular stiffness, the transmission and generation of higher contractile forces pathway and increased cytoskeletal remodeling dynamics. The focus of this study was on the mechanism that leads to higher invasiveness of cancer cells with high expression of the αvβ3 integrin.
The functional mechanisms promoting the invasion of cancer cells are only fragmentarily investigated, but there is a common agreement that biomechanical factors may determine the speed of cell migration in dense 3D ECMs [40-41, 43, 46]. Among these factors are adhesion forces, degradation processes of the ECM through secretion of matrix-degrading enzymes, remodeling dynamics of the cytoskeleton, cellular stiffness and fluidity, and contractile force transmission and generation [39,43,[48][49][50]. To investigate which of these biomechanical factors contribute to the higher invasiveness of cells with high αvβ3 integrin expression that were isolated from transfected cell lines with high and low αvβ3 integrin expression from MDA-MB-231 breast carcinoma cells. In each case, cells with high αvβ3 integrin expression were highly invasive. The knock-down of the β3 integrin subunit using β3 integrin subunit specific siRNA dramatically decreased their invasiveness into 3D ECMs.
Integrins are known to be involved in cell adhesion, transmigration and invasion processes and in coupling of the actomyosin cell cytoskeleton to the microenvironment. The activation of integrin receptors could be through conformational changes after ligand binding and possibly through biomechanical stimulation [17,50]. The activation of integrins may be regulated through either increased affinity to ligands by enhancing the number of activated integrins and total integrins on the cell's surface or integrin translocation into lipid rafts, whereas this has only been reported for β1 integrin subunits [50]. However, the αvβ3 integrin has been reported to be associated with receptor tyrosine kinases such as the VEGF-2 receptor to promote cell motility [52]. The latter was not addressed in this study, but the first was preliminarily reported not to be crucial because alternations in the expression levels of the β3 integrin were observed in aggressive, malignant tumors, but not in benign [34,51]. Consistently, the knockdown of the β3 integrin subunit in αvβ3 high cells reduced their invasiveness into 3D ECMs and impaired the mechanical alternations indicating that the αvβ3 integrin active function increases the invasiveness of cancer cells into connective tissue. Taken together, these results indicate that the αvβ3 integrin is a key player in facilitating cancer cell invasion into 3D ECMs by regulating the biomechanical properties of invasive cancer cells.
The signal transduction pathways that connect integrin adhesion events with cellular stiffness, cytoskeletal remodeling dynamics, traction force generation and cell invasion are still elusive, although important components of other integrin receptors have been studied in detail, such as the activation of α5β1 integrins by ECM ligands [50,53], the formation of focal adhesions following biochemical integrin activation [54], the connection between focal adhesion assembly and contractile forces [50,[55][56], and between contractile forces and 3D cell invasion [39,57]. Here, this article demonstrates that increased expression of αvβ3 integrins leads to increased cellular stiffness and cytoskeletal remodeling dynamics that enable aggressive cancer cells to overcome the steric hindrance of dense 3D ECMs. These findings are consistent with our previous studies that showed increased cancer cell stiffness and/or increased contractile forces of highly invasive cancer cells expressing high amounts of α5β1 integrins or CXCR2 receptors compared to the weakly invasive cancer cells expressing low amounts of α5β1 integrins or CXCR2 receptors [8,39]. The results of this study are in line with the other studies [8,39] and lead to the suggestion that several biochemical alterations such as integrins or chemokine receptors on the cellular surface of cancer cells may affect biomechanical properties in a similar way.
Blocking of the myosin contraction through the MLCK inhibitor ML-7 or the ROCK inhibitor Y27632 diminished the invasiveness of αvβ3 high cells indicating that the αvβ3facilitated increased invasiveness is contractile force dependent. In addition, both contractility inhibitors reduced the stiffness of αvβ3 high cells. These findings indicate that the αvβ3 integrin facilitated invasiveness into 3D microenvironments depends on the cellular stiffness and subsequently on the transmission and generation of contractile forces. However, it remains elusive whether increased stiffness caused increased contractile forces or vice versa. What is known is that cytoskeletal remodeling affects cellular stiffness and contractile force transmission and generation. In αvβ3 low cells neither the inhibitor Y27632 nor the ML-7 inhibitor reduced the invasiveness or stiffness, whereas in αvβ3 high cells the invasiveness and stiffness were significantly reduced indicating that αvβ3 high cells can employ invasion strategies that rely on the cellular stiffness and the generation of contractile forces. Nonetheless, by increasing the contractility of αvβ3 low cells through myosin light chain phosphorylation using Cal A, they showed increased invasion. These results indicate that contractile forces are needed for increased cell invasion. Furthermore, the inhibition of the Src kinase and its downstream targets STAT3 and Rac1 reduced significantly the invasiveness of αvβ3 high cells, but had no effect on the invasiveness of αvβ3 low cells. These results suggest that Src kinase/STAT3/Rac1 signaling pathways may be involved in the αvβ3 facilitated invasiveness of cancer cells. Mutational alteration of the Src-binding motif of p130Cas has reported to abolish the interaction between Src and its substrate p130Cas [58][59] and significantly reduce levels of p130Cas phosphorylation at this site [60][61]. The ability of Src to phosphorylate p130Cas is enhanced by mechanical extension of the interaction site, implying that p130Cas acts as a mechanosensor [62] or as indicated in this study STAT3 or RAC1 act as a mechanoregulatory proteins. In addition, p130Cas phosphorylation has been detected predominantly in nascent integrin adhesion sites [60,63], which are necessary for cell invasion into 3D ECMs and contractile force transmission and generation. All three inhibitors (Src, STAT3 and RAC1 inhibitors) reduced the stiffness of αvβ3 high cells, but not of αvβ3 low cells indicating that the αvβ3 integrin facilitated cancer cell invasiveness in a 3D microenvironment such as a 3D ECM is stiffness and hence, contractile force dependent.
Finally, it can be concluded that the cellular stiffness and subsequently the transmission and generation of contractile forces together with enhanced cytoskeletal remodeling dynamics are the driving factors for increased the αvβ3 integrin-facilitated cell invasion into 3D ECMs, and proposed that the measurement of biomechanical properties may be a novel factor in determining and explaining the malignancy of tumors.

Isolation of subcell lines
The human breast cancer cell line MDA-MB-231, which expresses endogenous αvβ3 integrins, was used for the generation of subcell lines with high (αvβ3 high cells) and low endogenous expression of αvβ3 integrins (αvβ3 low cells). To isolate high and low expressing subcell lines MDA-MB-231 cells were stained with an mouse antibody against the human αvβ3 integrin (clone LM609, Millipore) for 30 min at 4 • C and then, with a secondary R-PE-labeled goat anti-mouse-IgG (F(ab) 2 -fragment; Dianova) antibody for 30 min at 4 • C. Then, single cells of both subgroups (high and low) were sorted in ten 96 well plates using a cell sorter and grown to colonies and finally cultured to subcell lines at 95% humidity, 37 • C and 5% CO 2 in an incubator. Subcell lines high and low endogenous αvβ3 integrin expression were analyzed at least three times for each clone and remained stable over the whole investigation time (data not shown). Additionally, αvβ3 integrin subcell lines were isolated from T24, A375 and 786-O cancer cells with endogenous αvβ3 integrin expression [8,29].

3D ECM invasion assay
A 3D-collagen-invasion assay was used to study invasiveness of cancer cells. In particular, for a six well plate 3.5 ml collagen R (Serva, Heidelberg, Germany) and 3.5 ml collagen G (Biochrom, Berlin, Germany) were mixed, 0.8 ml of 278 mM sodium bicarbonate (end concentration 26.5 mM) and 0.8 ml 10 × DMEM (Biochrom) was added. Air bubbles were avoided during the 3D-ECM preparation process [8]. The polymerization of the 3D ECMs started after the stirred solution was neutralized with 1 N sodium hydroxide and incubated at 37 • C, 95% humidity and 5% CO 2 . Immediately, 1.2 ml of the collagen solution was pipetted into each well of a six-well plate, and collagen mixtures were polymerized. Polymerized 3D ECMs were normally 500 µm thick. The 3D collagen-matrices were incubated overnight with 2 ml DMEM [39]. After removal of the incubation medium, 100 000 cancer cells were seeded on top of the 3D ECMs and cultured for 72 h at 37 • C, 5% CO 2 and 95% humidity in DMEM containing 10% FCS. At this time period, differences in the invasiveness of cells were clearly visible. For serumfree cell invasion, cancer cells were cultured 24 h before and during the invasion assay in EX-cell293 medium (SAFCBiosciences, Lenexa, KS) with 100 U ml −1 penicillin-streptomycin. After 3 days, cancer cells cultured on and inside the collagen matrices were fixed with 2.5% glutaraldehyde solution in PBS-buffer, the fraction of cancer cells that invaded 3D ECMs and their invasion depth were measured by optical sectioning in 12 randomly selected fields of view. After this time, differences in the invasiveness of cell lines were clearly visible. Noninvasive cells can be readily identified by their nuclei (stained with 1 µg ml −1 Hoechst 33342 dye) located in one layer that coincides with the location of the topmost collagen fibers. A cell was counted as invasive when its nucleus is located below the layer formed by the non-invasive cells. Because of the depth of field of a 40 × 0.6 NA objective, the uncertainty of this method is ∼5 µm. The invasion depth was determined by focusing the microscope to the center of the nucleus; the value was read from the motorized z-drive of the microscope and was corrected for the refractive index of water (1.33). The z-focus at the gel surface served as the reference.
To determine the percentage of invasive cancer cells (cells inside the 3D ECMs), the adherent cells on top of the 3D collagen matrices were also counted [8]. The invasion profile plots only contain the invasive cancer cells. The percentage of cancer cells that invaded is given in the bar graphs.

Migration speed in 2D
10 000 αvβ3 low or αvβ3 high cells were seeded on FN or collagen type I coated glass slides. Cell movements were computed from phase-contrast images recorded with 10× magnification using a Fourier-based difference-with-interpolation algorithm [64]. The MSD of cell movements with time t was described with a power-law relationship MSD = D(t/t 0 )β [65][66][67], where t 0 is the time interval of the image recordings (1 min), D is the apparent diffusion coefficient and the power-law exponent β is a measure of persistence, with β ∼ 1 for randomly and β ∼ 2 for ballistically migrating cells [64]. The average migration speed over any time period can be obtained from the square root of the MSD and is shown in the figure. The MSD of cancer cells reveals that the migration process is not a Brownian random walk but is superdiffusive due to directional persistence and temporal fluctuations in migration speed.

Enzyme-linked immunosorbent assay
Supernatant from 100 000 cells of each subcell line was collected after 3 days of cell invasion and the FCS was measured. Fibronectin concentrations were determined using a Fibronectin-Elisa Kit according to the manufacturer's instructions (R&D systems).

Magnetic tweezer
Magnetic tweezers were used to apply a staircase-like sequence of step-forces ranging from 0.5 to 10 nN to superparamagnetic epoxylated 4.5 µm beads, coated with 100 µg ml −1 fibronectin from human plasma (Roche Diagnostics, Mannheim, Germany, Cat. No. 11080938001) or 100 µg ml −1 collagen type I (Biochrom) [39,49]. Fibronectin was present in 3D-ECMs, because it was embedded into the matrices from the 10% FCS in the cell culture medium as well as from cultured cancer cells, which secrete fibronectin. 2 × 10 5 fibronectin coated beads were sonicated, added to 10 5 cells, and incubated for 30 min at 37 • C and 5% CO 2 . All measurements were performed at 37 • C with an inverted microscope (DMI-Leica) after at least 30 min of bead binding, which is sufficient to connect the coated beads to the cytoskeleton of cancer cells. Once the coated beads are firmly connected to the cytoskeleton of cancer cells, the molecular details of the connection matter little and have no impact on the motion of coated beads [8]. The creep response J(t) of cells during force application followed a power law in time, J (t) = a(t/t 0 ) b , where the pre-factor a and the power-law-exponent b were force-dependent, and the reference time t 0 was set to 1 s. The bead displacement in response to a staircase-like force pattern followed a superposition of power laws from which the force dependence of a and b was determined by a least-squares fit [49,68]. The parameter a (units of µm nN −1 ) characterizes the elastic cell properties and corresponds to a compliance (the inverse of stiffness) [49]. The force-distance relationship in units of nN µm −1 is related to cellular stiffness in units of Pa by a geometric factor that depends on the contact area between the coated bead and the cancer cell (or the degree of bead internalization), and the cell height. If those parameters are known, the geometric factor can be estimated from a finite element analysis [69]. Without knowledge of the cell height and coated bead internalization, the typical strain ε can be estimated as the bead displacement d divided by the bead radius r, and the typical stress σ as the applied force F divided by the bead cross-sectional area πr 2 such that the cell stiffness corresponds to: G = σ/ε = r/d F/(πr 2 ) [70]. For 4.5 µm coated beads as used in this study, the geometric factor is 0.14 µm −1 , and a cell with an apparent stiffness of 1 nN µm −1 would have a 'proper' stiffness of 140 Pa.
The power-law-exponent b reflects the stability of force-bearing membrane and cytoskeletal structures connected to coated beads. Values of b = 1 and 0 indicate Newtonianviscous and elastic behavior, respectively [68]. A non-zero power-law exponent indicates that an amount of the deformation energy during magnetic force application is not elastically stored in the cytoskeleton, instead, it is dissipated as heat due to the remodeling of cytoskeletal structures to which the bead is bound [68]. Thus, the dissipation is directly coupled to the rate at which the elastic bonds within the cytoskeletal network break up and finally turn over. The turnover of acto-myosin bonds contributes to the dissipative properties [71], and is not considered as a remodeling event, but it enables contractility-driven morphological shape changes of the cytoskeleton. In cancer cells, the power-law exponent b usually falls in the range between 0.1 and 0.5, whereby higher b-values have been linked to a higher turn-over rate of cytoskeletal structures [72]. The aand b-values are averaged over all applied forces and all coated beads measured (bound to cells) and are expressed as mean ± SE.

Statistical analysis
The flow cytometry data, 2D migration data and enzyme-linked immunosorbent assay data were expressed as mean values ± SD. All other data were expressed as mean values ± SE. Statistical analyses were performed using the two-tailed Student's t-test. A value of p < 0.05 was considered to be statistically significant.